Investigating Impact of New Information

As mentioned in section 4.3.6, new information regarding the characteristics of the modules were discovered. The change in module characteristics consists of changing the shunt resis-tance ofR_sh=250≠and series resistanceRs=0,13 ≠and the short circuit current tempera-ture coefficient to I_sctc =2,2m A/^±C. Simulating energy production with default loss settings as described in scenario 2 (S2) resulted in a yearly energy estimate of 286036 kW h. That is a underestimate of 4,64 % compared to produced energy and more than a 4 percentage points lower estimate compared to scenario 2. The resulting monthly deviation is 2606,1kW h, which is about a 58 % increase of monthly deviation from S2. There are two aspects that have to be an-alyzed from this result. The first is the reason for the increased underestimate from S2 (greater deviations) and the second is how it may impact the other results.

Performance ratio drops from 82,8 % in S2 to 79,3 %, indicating increased losses. The loss di-agram for the combined scenario is presented in appendix B.4. The loss that emerge from the diagram is the notable increase in PV loss due to irradiation level. The loss increased 3,9 per-centage points from S2, which constitutes most of the total underestimate of 4,64 %. PV loss due to irradiation level is a result of the intrinsic behavior of the solar cell, described by the one diode model (2.7), and is thus mostly determined by the module resistances. Most of the reaming loss is due to increased PV loss due to temperature, a result of the decrease ofIsc_tc. Ac-cording to PVsyst (2014) modules with lowR_sh and highR_s have the best performances under low light conditions. Due to the increased loss, it appears that the decrease ofR_s influence the results more than the decrease ofRsh. The decrease ofRsh should ideally contribute to better low light performance, as Rsh increases exponentially when irradiation is decreased. As illus-trated in figure A.10b, the increasedR_shcontributes to a higher gap in saved efficiency, however, mostly at irradiation levels below 400W/m. Figure A.10a shows how a higherRsmaintains high efficiency for lower irradiation levels compared toR_sof lower values. Adjustment of module re-sistances can occur with upgrades of PVsyst and as a result change loss due to irradiation level, as reported by PVsyst users at the PVsyst forum Mermoud. That is mainly due to the manner in which the resistance are estimated by PVsyst is changed (PVsyst (2014)). As some estimation models in PVsyst changes with an update, PVsyst should not be updated during a project. These changes can be hard to track.

Analyzing the impact of adjusting the module resistances in PVsyst proved the above theories.

DecreasingR_shslightly increased yearly estimated energy, while decreasingRsresulted in a

sig-nificant decrease. The sigsig-nificant decrease in simulated energy with decreasedRsexplains why the original module resistances performs better than the alternative characteristics. This is in-teresting in two manners. Firstly, it implies that the impact of reducingRs in Norway reduces produced energy. At STC conditions, lowRs resistance performs better than high. High irradi-ation levels results in high currents, which again results in high power losses (eq. 2.6 and eq.

2.10). That indicates that most of the received irradiation at ASKO must be much lower than the 1000W/m²(STC). Low-light performance measurements of the module included in PVsyst can thus improve accuracy of the simulation, as PVsyst will be more able to adjust the series resis-tance accordingly. The second, is the impact of the shunt resisresis-tance. Studies done by Bunea et al. (2006) and Reich et al. (2009) argue that a higher shunt resistance will retain more of its efficiency at lower irradiation levels, while low shunt resistances will decrease linearly with irra-diation.

For the previous results this implies an uncertainty, and that the results may be reduced with the proven underestimate to account for the additional losses presented in this section. The additional underestimate applies to scenario 1 (S1) and S2, as well as the adjustment of each parameter based on S2. That is because the factors causing the changes lie within the modules, which are used in all simulations. The increased underestimate corresponds better compared to previous studies, as described in the introduction (1). Yet, the coherence between simulated and produced energy is more accurate than the results of previous studies. The patterns described when adjusting individual parameters still apply, although deviation is now more significant.

The optimized scenario (Comb-1) shows a considerable overestimation. This overestimation is compensated for, when the new information is applied. With the new information PVsyst un-derestimates simulated energy more compared to S2. That implies that all meteo sources in S1 and S2 will underestimate even more than what is shown in the result. The overestimation of the Comb-1 scenario is to some extent neutralized with further underestimation of the original S2 result. A new combined scenario (Comb-3) can be simulated based upon the exact same de-tailed loss parameter as in Comb-1, except that the newR_sh,RsandIsc°Temper atur e°Coe f f i ci ent

is updated for the module characteristics.

The Comb-3 scenario resulted in a yearly energy estimate of 297713kW h, only a 0,75 % under-estimate of actual energy. The performance ratio increases to 83,1 %, indicating that the losses are reduced for the combined scenario. The Relative Error (RE) for the new combined scenario (Comb-3) is presented in figure 5.17. The RE for the Comb-3 scenario shows a significant

un-Figure 5.17: Relative Error for the Comb-3 scenario. This scenario is based upon the Comb-1 scenario with the only difference being the adjustment of Rsh, Rs and Isc°Temper atur e°Coe f f i ci ent.

derestimate during winter months. The summer RE values are all below 1 %. The experimental soiling values used in the combined scenarios aimed to reduce the considerable overestimate during winter moths. Further reduction caused by the new module characteristics, as proved for S2 above, results in the negative RE values. Due to the nature of the one diode model, the result is more evident during winter (higher RE) due to lower light conditions. The Root Mean Square Deviation (RMSD) for the Comb-3 scenario is 317,28kW h, which is a 80,74 % reduction of monthly deviation from S2. Although it is not lower than the RMSD in Comb-2, it does not include the LID factor and increased summer soiling levels. The Comb-3 scenario presents a strong positive correlation with a R-squared value of 99,988 %, as shown in figure 5.18.

Further adjusting the experimental soiling levels to reduce the underestimation in Comb-3 for winter months improves the accuracy of the simulation drastically. Decreasing the winter soil-ing levels, as described in appendix B.3, results in monthly RMSD of 167,29kW h. That is half the monthly deviation of the Comb-2 RMSD, and a corresponding R-squared value of 99,993 %, presenting a very strong correlation between the simulated and produced energy. Thus, half the

Figure 5.18: Correlation plot for the Comb-3 scenario. The plot shows a strong positive correla-tion. The R-squared value is 99,988 %.

deviation in Comb-3 was due to underestimation of predicted energy during winter, which was a result of the experimental soiling values.

The results of this section is also rather interesting. Input parameters and characteristics of modules are as important for the simulation estimate as simulations parameters (detailed losses).

A change in module resistance toR_sh=250≠andRs=0,13≠reduced yearly simulated energy over 4 %. The results presented in this chapter shows that albedo and ohmic adjustment influ-ence simulated energy relatively small compared to aggressive soiling values or thermal losses according to module temperature. That implies that quality of data used when simulating is just as important as which detailed losses are included and how much the impact. For example is the yearly impact of changing theRsh andRsmore significant than adjusting soiling values.

5.5.1 The combined scenarios on the Meteonorm data

The Comb-1 and Comb-3 scenarios are applied with the Meteonorm (MN) meteo file to inves-tigate how adjustments of the detailed loss parameters work on the other meteo files. These scenarios are named MN-Comb-1 and MN-Comb-3. Since MN does not included module tem-perature, thermal loss according to own defined heat loss constants are used,Uc=25 (W/m²K) andUv =1,4 ((W/m²K/(m/s)). The detailed loss parameters in Comb-1 and Comb-3 are ex-actly the same. The only difference is the update on module characteristics.

The MN-Comb-1 resulted in a yearly simulation estimate of 293328 kW h , which is a 2,2 % underestimate of actual energy production. Although the underestimate is about the same for MN-Comb-1 and MN meteo in scenario 2 (S2), the monthly RMSD decreased to 3118,6kW h (about a 20 % reduction). Compared to the RMSD for Comb-1 with ASKO meteo, this is almost 3 times the deviation. Yearly performance ratio was estimated to 83,4 %, which is not that much lower than combined parameters for ASKO meteo data. That suggests that expected energy pro-duction at STC is lower for MN meteo than for ASKO meteo. The monthly Relative Error (RE) is presented in figure 5.19. The figure presents high RE´s, with both under- and overestimations.

The results from the MN-Comb-1 is distinct from the Comb-1 result in the way adjusting the parameters impacted the simulation results. The comb-1 scenario resulted in a significant in-crease in predicted yearly energy from S2 for ASKO meteo. However, for MN meteo, the yearly result is about the same.

It is uncertain why the adjustment of detailed loss parameters does not impact the simulation result in a similar manner for MN meteo as for ASKO meteo. By analyzing the loss diagram for MN-Comb-1 it is clear that some of the loss factors are increased compared to Comb-1 for ASKO meteo. This may be due to how the meteo data impact the individual loss parameters.

MN recorded higher wind values, which should contribute to a higher thermal heat loss (in-creased predicted energy), compared to ASKO meteo. The estimated ambient temperature by MN is higher than ÅS meteo contributing to a decrease in thermal heat loss. Although yearly irradiation is about the same for ASKO and MN meteo, monthly irradiation by MN is less in June and July, a pattern seen in figure 5.19.

The alternative module characteristics in the MN-Comb-3 scenario resulted in a yearly simula-tion of 282277kW h, which is a underestimate of 5,9 % of actual energy production. Considering the accuracy of the Comb-3 for ASKO meteo, this scenario resulted in a significant

underesti-Figure 5.19: Relative Error for MN-Comb-1 and MN-Comb-3 scenarios compared to produced energy at ASKO.

mate. Monthly RMSD was 3524,9kW h, which is an increase from the MN-Comb-1 scenario.

Figure 5.19 presents the decrease in predicted energy. The performance ratio drops from the MN-Comb-1 to 80,3 %, indicating increased losses. It is clear that with the alternative module characteristics, MN meteo underestimate considerably. The underestimate seen with the alter-native module characteristics correlated better with previous studies, as described in section 1.

The results presented for the combined scenarios with MN meteo data implies that satellite col-lected data underestimate significantly more than meteo data colcol-lected from weather stations, as ASKO. MN underestimated the most for S2, and combining the parameters did not improve the simulation result notably. The validity of introducing a possible LID loss and increasing summer soiling levels in the Comb-2 scenario with ASKO meteo data, can be discussed consid-ering the underestimate with alternative module characteristics. Comb-3 with ASKO meteo and MN-Comb3 both resulted in a a underestimate. The underestimate in MN-Comb-1, implies that there is no reason to introduce these losses in a Comb-2 scenario with MN meteo data.

Section summary

New module characteristics reduced yearly simulated energy and performance ratio signifi-cantly, mostly due to a increase in PV loss due to irradiation level. The impact will be evident in all presented results with ASKO meteo. The alternative module characteristics are not confirmed by the manufacturer after it was discovered a difference within their two data sheets. Satellite collected data by MN underestimates considerably more compared to measured meteo data at ASKO. Adjustment of detailed loss parameters influence ASKO simulation results more than MN simulations result. Decreasing series resistance significantly decreases yearly simulated energy.

The combined scenarios question the introduction of a LID loss, and supports low summer soil-ing levels (1 %). Furthermore, it implies lower experimental soilsoil-ing levels dursoil-ing winter.